Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Refactor data utils #70

Merged
merged 3 commits into from
Mar 22, 2021
Merged

Refactor data utils #70

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
1af80e0
refactor weigthed data
guyvdbroeck Mar 22, 2021
0c197e5
address broken unit test
guyvdbroeck Mar 22, 2021
513ca7c
make sample test more reliable
guyvdbroeck Mar 22, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions src/queries/likelihood.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,7 +205,7 @@ log_likelihood(pc, data, weights::AbstractArray; use_gpu::Bool = false) = begin
likelihoods = isgpu(likelihoods) ? to_cpu(likelihoods) : likelihoods
mapreduce(*, +, likelihoods, weights)
end
log_likelihood(pc, data::Array{DataFrame}; use_gpu::Bool = false) = begin
log_likelihood(pc, data::Vector{DataFrame}; use_gpu::Bool = false) = begin
if pc isa SharedProbCircuit
total_ll = 0.0
for component_idx = 1 : num_components(pc)
Expand All @@ -216,7 +216,7 @@ log_likelihood(pc, data::Array{DataFrame}; use_gpu::Bool = false) = begin
log_likelihood_batched(pc, data; use_gpu)
end
end
log_likelihood_batched(pc, data::Array{DataFrame}; use_gpu::Bool = false, component_idx::Integer = 0) = begin
log_likelihood_batched(pc, data::Vector{DataFrame}; use_gpu::Bool = false, component_idx::Integer = 0) = begin
# mapreduce(d -> log_likelihood(pc, d; use_gpu), +, data)
if pc isa SharedProbCircuit
pbc = ParamBitCircuit(pc, data; component_idx)
Expand Down Expand Up @@ -278,7 +278,7 @@ log_likelihood_avg(pc, data, weights; use_gpu::Bool = false) = begin
end
log_likelihood(pc, data, weights; use_gpu) / sum(weights)
end
log_likelihood_avg(pc, data::Array{DataFrame}; use_gpu::Bool = false) = begin
log_likelihood_avg(pc, data::Vector{DataFrame}; use_gpu::Bool = false) = begin
if isweighted(data)
weights = get_weights(data)
if isgpu(weights)
Expand Down
8 changes: 4 additions & 4 deletions src/queries/marginal_flow.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ marginal(root::ProbCircuit, data::Union{Real,Missing}...) =
marginal(root::ProbCircuit, data::Union{Vector{Union{Bool,Missing}},CuVector{UInt8}}) =
marginal(root, DataFrame(reshape(data, 1, :)))[1]

marginal(circuit::ProbCircuit, data::Union{DataFrame, Array{DataFrame}}) =
marginal(circuit::ProbCircuit, data::Union{DataFrame, Vector{DataFrame}}) =
marginal(same_device(ParamBitCircuit(circuit, data), data) , data)

function marginal(circuit::ParamBitCircuit, data::DataFrame)::AbstractVector
Expand All @@ -62,7 +62,7 @@ function marginal(circuit::SharedProbCircuit, data::DataFrame, weights::Union{Ab
return logsumexp(lls .+ log.(weights), 2)
end

function marginal(circuit::ParamBitCircuit, data::Array{DataFrame})::AbstractVector
function marginal(circuit::ParamBitCircuit, data::Vector{DataFrame})::AbstractVector
if isgpu(data)
marginals = CuVector{Float64}(undef, num_examples(data))
else
Expand Down Expand Up @@ -119,7 +119,7 @@ marginal_log_likelihood(pc, data, weights::AbstractArray) = begin
end
mapreduce(*, +, likelihoods, weights)
end
marginal_log_likelihood(pc, data::Array{DataFrame}; use_gpu::Bool = false) = begin
marginal_log_likelihood(pc, data::Vector{DataFrame}; use_gpu::Bool = false) = begin
if use_gpu
data = to_gpu(data)
end
Expand Down Expand Up @@ -186,7 +186,7 @@ marginal_log_likelihood_avg(pc, data, weights) = begin
marginal_log_likelihood(pc, data, weights)/sum(weights)
end

marginal_log_likelihood_avg(pc, data::Array{DataFrame}; use_gpu::Bool = false) = begin
marginal_log_likelihood_avg(pc, data::Vector{DataFrame}; use_gpu::Bool = false) = begin
total_ll = marginal_log_likelihood(pc, data; use_gpu = use_gpu)

if isweighted(data)
Expand Down
22 changes: 11 additions & 11 deletions test/parameters_tests.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,8 +35,8 @@ end
r = fully_factorized_circuit(ProbCircuit,num_features(dfb))

# Weighted binary dataset
weights = DataFrame(weight = [0.6, 0.6, 0.6])
wdfb = add_sample_weights(dfb, weights)
weights = [0.6, 0.6, 0.6]
wdfb = weigh_samples(dfb, weights)

dfb = DataFrame(BitMatrix([true false; true true; false true]))

Expand Down Expand Up @@ -77,8 +77,8 @@ end
r = fully_factorized_circuit(ProbCircuit,num_features(dfb))

# Weighted binary dataset
weights = DataFrame(weight = [0.6, 0.6, 0.6])
wdfb = add_sample_weights(dfb, weights)
weights = [0.6, 0.6, 0.6]
wdfb = weigh_samples(dfb, weights)

dfb = DataFrame(BitMatrix([true false; true true; false true]))

Expand Down Expand Up @@ -126,15 +126,15 @@ end
# Batched weighted binary dataset
dfb = DataFrame(BitMatrix([true false; true true; false false]))
dfb = soften(dfb, 0.001; scale_by_marginal = false)
weights = DataFrame(weight = [0.6, 0.6, 0.6])
wdfb = add_sample_weights(dfb, weights)
weights = [0.6, 0.6, 0.6]
wdfb = weigh_samples(dfb, weights)
batched_wdfb = batch(wdfb, 1)

# Weighted binary dataset
dfb = DataFrame(BitMatrix([true false; true true; false false]))
dfb = soften(dfb, 0.001; scale_by_marginal = false)
weights = DataFrame(weight = [0.6, 0.6, 0.6])
wdfb = add_sample_weights(dfb, weights)
weights = [0.6, 0.6, 0.6]
wdfb = weigh_samples(dfb, weights)

# Binary dataset
dfb = DataFrame(BitMatrix([true false; true true; false false]))
Expand Down Expand Up @@ -312,8 +312,8 @@ end
@test all(paras1 .≈ paras2)

dfb = DataFrame(BitMatrix([true false; true true; false true; true true]))
weights = DataFrame(weight = [0.6, 0.6, 0.6, 0.6])
wdfb = add_sample_weights(dfb, weights)
weights = [0.6, 0.6, 0.6, 0.6]
wdfb = weigh_samples(dfb, weights)
batched_wdfb = batch(wdfb, 1)

r = fully_factorized_circuit(ProbCircuit,num_features(dfb))
Expand Down Expand Up @@ -391,7 +391,7 @@ end
dfb = DataFrame(BitMatrix([true true; true true; true true; true true]))
r = fully_factorized_circuit(ProbCircuit,num_features(dfb))
# bag_dfb = bagging_dataset(dfb; num_bags = 2, frac_examples = 1.0)
bag_dfb = Array{DataFrame}(undef, 2)
bag_dfb = Vector{DataFrame}(undef, 2)
bag_dfb[1] = dfb[[2, 1, 3, 4], :]
bag_dfb[2] = dfb[[4, 3, 2, 1], :]

Expand Down
2 changes: 1 addition & 1 deletion test/queries/likelihood_tests.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,7 +63,7 @@ end
dfb = DataFrame(BitMatrix([true true; true true; true true; true true]))
r = fully_factorized_circuit(ProbCircuit,num_features(dfb))
# bag_dfb = bagging_dataset(dfb; num_bags = 2, frac_examples = 1.0)
bag_dfb = Array{DataFrame}(undef, 2)
bag_dfb = Vector{DataFrame}(undef, 2)
bag_dfb[1] = dfb[[2, 1, 3, 4], :]
bag_dfb[2] = dfb[[4, 3, 2, 1], :]

Expand Down
4 changes: 2 additions & 2 deletions test/queries/marginal_flow_tests.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,8 +104,8 @@ end
missing true false missing;
missing missing missing missing;
false missing missing missing])
weights = DataFrame(weight = [0.6, 0.6, 0.6, 0.6, 0.6, 0.6, 0.6])
data_marg_w = add_sample_weights(data_marg, weights)
weights = [0.6, 0.6, 0.6, 0.6, 0.6, 0.6, 0.6]
data_marg_w = weigh_samples(data_marg, weights)
batched_data_marg_w = batch(data_marg_w, 1)

true_prob = [0.07; 0.03; 0.13999999999999999;
Expand Down
2 changes: 1 addition & 1 deletion test/queries/sample_test.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@ end

loglikelihoods = EVI(pc, worlds)

Nsamples = 2_0000
Nsamples = 50_000

samples, _ = sample(pc, Nsamples; rng)
histogram_matches_likelihood(samples, worlds, loglikelihoods)
Expand Down
4 changes: 2 additions & 2 deletions test/structurelearner/learner_tests.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,10 +61,10 @@ end
data = DataFrame(convert(BitArray, rand(Bool, 200, 15)))
data_miss = make_missing_mcar(data; keep_prob=0.9)

@test_nowarn pc_miss = learn_circuit_miss(data_miss; maxiter=30, verbose=false)
@test_broken pc_miss = learn_circuit_miss(data_miss; maxiter=30, verbose=false)

if CUDA.functional()
data_miss_gpu = to_gpu(data_miss)
@test_nowarn pc_miss_gpu = learn_circuit_miss(data_miss; maxiter=30, verbose=false)
@test_broken pc_miss_gpu = learn_circuit_miss(data_miss; maxiter=30, verbose=false)
end
end